基于背景重建和MDBP的视频对象分割算法

A Video Object Segmentation Algorithm Based on Background Construction and MDBP

提出了背景置信度图像和背景标示图像的概念,给出了一种基于背景重建和象素最小距离(M DBP)的自动视频对象分割方法。首先运用基于背景置信度图像和背景标示图像的背景重建技术从视频序列的多帧图像中重建出可靠的背景图像,然后运用差背景法分割视频对象(VO),同时再利用象素最小距离(M D BP)和总体象素最小距离(W M DBP)准则对分割出的视频对象图像进行处理,克服由于背景的微小变化而引起的前景对象的错误检测。试验结果表明该文给出方法能够较好地重建背景,对于背景静止的视频能够得到更好的分割结果。

In this paper we present concepts of background confidence image and background label image.An automatic background construction and minimum distance between pixels based video object segmentation algorithm is proposed in this paper.In this algorithm,firstly,a background confidence image and background label image based background construction technique is used to construct a reliable background image from multiple frames.Then segment the video object by abstract background.At the same time,the minimum distance between pixels(MDBP)and the whole MDBP rules are given in this paper to remove error detection of video object caused by small change of background.Experimental results show that this algorithm can obtain better construction background image,and get better segmentation result of the video that background is static.